As a data scientist, the time I served on a North Carolina campaign committee in 2018 was an eye-opener, revealing just how far behind the curve the state’s Democrats are when it comes to using data strategically to help win elections (a problem that I’m sure is not unique to our state). Thus was born, this month, EQV Analytics: a data science team devoted to helping the state’s Democratic candidates in 2020 and beyond. A key challenge facing EQV is first to educate those candidates regarding what the heck data science is, and how it can help them across the finish line. This is the first in a series of EQV white papers addressing those questions. Copyright 2019 EQV Analytics; reprinted by permission.
Unaffiliated voters: the X factor
Citizens of 29 U.S. states have the opportunity to choose a party affiliation when they register to vote, or else to register as 'unaffiliated.' Here in North Carolina those unaffiliated voters (UNAs) are the electorate's fastest-growing segment, outnumbering Republicans and on course to overtake Democrats (see Figure 1, above).
Because they don't wear their political sympathies on their sleeves, UNAs are the unknown X factor in political calculations. Pundits who consider them at all often do so only in generalizations that are short on actionable insights, such this recent exchange between North Carolina party officials:
“They choose not to be affiliated with one of the parties, but this doesn't make them any less partisan," [Republican director Dallas] Woodhouse said. [Democratic vice chair Matt] Hughes agreed: "They are very hardcore in their political beliefs."
But viewing unaffiliated voters as a monolithic mass of partisans without parties – particularly in the absence of solid evidence – can dangerously mislead campaigns into forgoing efforts to court the swing voters that must surely be numbered among them. Are swing-voting UNAs' numbers sufficiently great to deserve a progressive campaign's attention? Fortunately, here in North Carolina that question is uniquely answerable.
Solving For X
We can know (rather than guess) the political leans of unaffiliated Tar Heels because North Carolina is among the nation's few states with both semi-closed primaries and open voter history data. 'Semi-closed' means that UNAs are permitted to vote in the primaries of the party of their choice (while party-registered voters are permitted to vote only in their own party's primary). And 'open data' means that UNAs' primary choices are a public record.
EQV analysts recently completed a study of the voter histories of every North Carolina primary-voting UNA from 2010 through 2018. We first distinguished partisan-UNAs (unaffiliated voters who always choose the same primary party) from swing-UNAs (those who don’t). Next we categorized the latter's current affinities (swing-Dem UNA or swing-Rep UNA), employing a proprietary algorithm that considers both how often and how recently their primary voting choices have swung from one party to the other.
Slightly more than 50% of the 1+ million North Carolina UNAs who voted in the 2018 general election have primary voting histories sufficient to permit us to classify their party affinities. Here's their breakdown:
Classification |
Number
(1,000s)
|
Fraction |
Partisan UNAs: |
Partisan-Dem |
209 |
40% |
Partisan-Rep |
267 |
52% |
Swing UNAs: |
Swing-Dem |
25 |
5% |
Swing-Rep |
16 |
3% |
Our analysis confirms the common opinion that most unaffiliated voters (here, 92% of them) are indeed partisans without parties. Yet it also underscores the importance of working to win over the substantial minority of UNAs (8%) who are true swing voters – more than half of whom favored Democrats in 2018, for the first time in recent history (Figure 2).
Applying these same proportions to all UNAs who voted in 2018 (not just the half that our algorithm could classify) yields an estimate of about 81,000 swing-voting UNAs last year, roughly 50,000 of whom leaned Democratic. That’s a number well worth paying attention to in a battleground state like North Carolina, where races are often won or lost by very slim margins.
Our analysis likewise underscores a unique advantage of quantitative analytics over more common methods of profiling the electorate, such as opinion polling or exit polls. Unlike those purely statistical approaches, EQV's fact-based analytics yield useful voter lists – in this example, the identities of current swing voters – that campaigns can use to power highly targeted outreach efforts.
Demographics
Not surprisingly, partisan UNAs look much like registered members of the parties they favor: Democratic-leaning partisans are younger, far more racially diverse, and more likely to be women than are Republican-leaners. Swing-voting UNAs, on the other hand, differ from partisans in important ways that may provide important insights into their distinctive concerns.
Median Age:
Since at least 2012 (Figure 3), partisan-Rep UNAs are old and trending older, while partisan-Dem UNAs are young and getting younger (their median ages in 2018 were 58 and 49, respectively). In contrast, swing-Rep and swing-Dem UNAs are about equally elderly (medians 62 and 60, respectively), and recent years have seen relatively little change.
Race:
Again like their party-affiliated peers, partisan-Dem UNAs are far more likely to be people of color (28% of their numbers in 2018, and growing) than are partisan-Reps (5%) (Figure 4). Meanwhile, swing-Dem UNAs are much more likely to be white than are partisan-Dems, as unaffiliated swing voters of both affinities are converging toward 8% people of color. The take-away here is that swing voting among UNAs appears to be largely a non-Hispanic white phenomenon.
Note: as used here, the term "people of color" comprises the Board of Elections-mandated groupings Asian, African American, American Indian, Hispanic, people of two or more races, and 'other.' The category thus excludes only non-Hispanic whites. The small minority of voters who choose not to specify their races when they register to vote are excluded from this portion of our analysis.
Gender:
Partisan-Dem UNAs are more likely to be female than are partisan-Rep UNAs, and that hasn't changed much in recent years (Figure 5). But on the swing side, 2018's apparent sharp increase in female swing-Reps deserves comment. Women didn't suddenly swing Republican last year. Rather, 2018 witnessed an unusually large number of previously swing-Rep men who (according to our algorithm) flipped to partisan-Rep last year, leaving women over-represented among remaining swing-Reps. While right-wing men's political sympathies appear to be hardening in the age of Trump, female conservatives' attachment to the Republican party remains considerably more provisional.
Breaking Blue
In 2017, NC Republican Party executive director Dallas Woodhouse observed that "since 2010, Republicans have won unaffiliated voters in every general election." Our timeline analysis confirms the general outline of that claim, but also reveals an important caveat: since 2014, Republicans’ advantage among unaffiliated voters has been eroding, with their margin narrowing by 2 to 3 points per election cycle (Figure 6).
Of course, history does not dictate destiny. Continued effort to to whittle away unaffiliated voters' support from the Republican Party is clearly warranted. Progressives who find this incremental trend maddeningly slow should take some comfort from the observation that Donald Trump won North Carolina in 2016 by just 3 points – a margin that was erased in 2018 among the state's unaffiliated voters. That trend helped hand North Carolina Democrats victory in every statewide race last year.